Do UAV-Derived Boulder Dimensions Match Traditional Field Measurements at the Grind o’ da Navir, Shetland?

Coastal boulder deposits are found worldwide on rock coasts and can serve as a record of extreme wave events. Boulder dimensions are key parameters in commonly used hydrodynamic equations for reconstructing the extreme wave events which likely emplaced them, providing data that can inform coastal hazard assessments. Traditionally, boulder dimensions are measured in the field using a tape measure, a process that can be time-consuming when collecting large datasets. Remote sensing approaches are increasingly being used in the measurement of coastal boulder deposits to extract boulder parameters. Orthomosaics generated from unmanned aerial vehicles (UAVs) allow for individual boulders to be measured using either manual or automated digitisation techniques. By digitising the outlines of boulders, a minimum bounding box can be fitted, and parameters such as orientation and axis lengths extracted. Large, site-wide datasets of these values can be rapidly generated, enabling whole site characterisation. However, this technique has only seen limited validation, particularly at sites with clustered boulders and an uneven basement surface. This study compares UAV-derived measurements of boulder dimensions with traditional field measurements, to test the statistical similarity.

The Grind o’ da Navir, Shetland, has an abundance of demonstrably storm wave emplaced boulders, which form ridges across the 15 – 20 m high cliff top. The site was selected due to its uneven bedrock and complex boulder ridge morphology, providing a boulder-abundant but methodologically-challenging environment. From 7950 digitised boulders, the mean long (A) axis was 0.66 m and the intermediate (B) axis was 0.40 m, with maximum axis lengths of 3.37 m and 1.73 m respectively. When compared using a paired subset of the data, the digitisation and traditional methods have a mean difference in A-axis of 0.05 m, and of 0.02 m for the B-axis, with a standard deviation of 0.14 m and 0.12 m for the two axes, respectively. This shows that the two methods are broadly interchangeable for average statistics with only minor bias, but individual measurements may have a larger error. The dip angle of boulders within the ridges at the Grind likely contributes to these individual measurement errors. At sites with isolated boulders on flat platforms, such discrepancies would be expected to be considerably smaller. Thus, UAV-derived boulder outlines can generate site-wide boulder statistics more rapidly than traditional field methods, with reasonable accuracy. The digitisation method can complement traditional field techniques, enabling larger, spatially extensive datasets while reducing the likelihood that spatial variability in boulder characteristics is overlooked.